A method and system for hierarchical generation of urban road network considering intersection turning relationship

Abstract

The application discloses a kind of urban road network grading generation method and system considering intersection turning relationship, comprising: step 1, different grade road section identification: trajectory kernel density grading and morphological refinement are used to obtain different grade road section information;Step 2, different grade road section merging and optimization: different grade road sections are merged, and topological construction is carried out to form a complete road network;Step 3, intersection turning relationship detection: based on intersection influence area, intersection associated road section detection is carried out, then the traffic condition between associated road sections is distinguished, and finally the turning relationship of the associated road section is identified, and the intersection turning relationship detection is completed.The present application generates different level road network based on hierarchical form, which can greatly ensure the accuracy of road network.The proposed intersection turning method is simpler and more clear than the previous method, and the turning relationship is determined based on the effective elimination of noise turning, which can effectively improve the identification accuracy.

Description

Technical Field

[0001] This invention belongs to the field of road network generation, and in particular relates to a hierarchical generation method and system for urban road networks that takes into account the turning relationships at intersections. Background Technology

[0002] In an increasingly urbanized society, cities are constantly expanding outwards, leading to continuous changes in the roads running through them and necessitating real-time updates to the urban road network. Traditional methods for acquiring updated road information are extremely costly in terms of resources, time, and accuracy, making it difficult to guarantee the real-time nature of the road network. Therefore, this paper employs readily available and low-cost GPS (Global Positioning System) trajectory data for road network extraction research.

[0003] Existing methods for road network identification include grid-based methods, incremental methods, clustering methods, and intersection connection methods. Among these, the grid-based method can effectively handle trajectory quality uncertainties and is applicable to trajectory data with low sampling frequencies. However, this method mostly generates the entire road network, failing to guarantee the precise geometry and correct topology of the road network. Road network identification is inseparable from the detection of intersection turning relationships. The effectiveness of these turning relationships directly determines the quality of urban road network navigation information and is of paramount importance in navigation information calculation. Currently, more and more researchers are conducting research on turning relationships based on crowdsourced trajectory data, providing some research ideas for intersection turning identification. However, existing clustering and map matching methods focus on judging turning trajectories and cannot effectively set thresholds to remove noisy turns. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a method for hierarchical generation of urban road networks that takes into account the turning relationships at intersections, comprising the following steps:

[0005] Step 1, Identification of different levels of road segments: Based on geometric interval classification, the kernel density of trajectory points is classified into different levels to achieve the division of kernel density of trajectory points of different levels. Then, a morphological refinement algorithm is used to refine the kernel density of trajectory points of different levels, and then different levels of road segments are generated.

[0006] Step 2, Merging and Optimizing Road Segments of Different Levels: Merge the identification results of road segments of different levels to obtain road segment node information. Then, based on this, perform effective identification of intersections using Delanuary triangulation clustering. Subsequently, conduct road network topology checks based on the intersections to achieve topological connection of broken road segments and form a complete road network.

[0007] Step 3, Intersection Turning Relationship Detection: Based on the identified intersection, the influence area of ​​the intersection is determined, and then the related road segments of the intersection are detected. Subsequently, the traffic conditions between the related road segments are judged, and finally, the turning relationship of the related road segments is identified by the turning rules, thus completing the intersection turning relationship detection.

[0008] Furthermore, the identification of different road segment levels mentioned in step 1 specifically includes:

[0009] Step 1.1: Perform quality cleaning on the trajectory data, including location duplication removal, stop point detection, and Delanuary triangulation denoising. Then, perform kernel density analysis on the cleaned trajectory and classify the trajectory kernel density according to geometric interval classification to achieve different levels of trajectory point kernel density division. The default division is three levels.

[0010] Step 1.2 involves morphological refinement of the kernel density of different levels of trajectory points, yielding the refinement results for first-level, second-level, and third-level trajectory point kernel densities.

[0011] Step 1.3: Identify the first-level trajectory point kernel density refinement results as first-level road segments, then establish a suitable buffer zone based on the first-level road segments to erase the second-level trajectory point kernel density refinement results to obtain second-level road segments. Finally, establish a buffer zone based on the first and second-level road segments to erase the third-level trajectory point kernel density refinement results to obtain third-level road segments.

[0012] Furthermore, the merging and optimization of road segments of different levels mentioned in step 2 specifically includes:

[0013] Step 2.1: Construct a Delaunay triangulation network based on the start and end points of the identified road segment, and use this network for intersection identification; the specific implementation method is as follows:

[0014] (a1) First, based on the node-connected road segments, the edges of the node-connected triangular network are removed, and the two triangular network edges with the largest angle to the road segment are retained.

[0015] (a2) Then, the edges of the remaining Delaunay triangulation are superimposed on the M-meter buffer of the identification road network. If the ratio of the overlap length to its length is less than a certain threshold, it is discarded.

[0016] (a3) Then, traverse each node, cluster the continuous nodes connected by triangles into one class, and remove the points that are not connected by triangles as noise until all nodes have been traversed once, and complete the cluster analysis of the nodes. The average value of the node position of each cluster is identified as the candidate position of the intersection.

[0017] (a4) To optimize the candidate intersection location, the candidate intersection location is adjusted to the point with the highest trajectory kernel density within the maximum range of distances from the candidate intersection to its corresponding cluster node, thus obtaining the final intersection location;

[0018] Step 2.2: Based on the intersection identification results and the clustering node information at the intersection, the information of the broken road segments connecting the intersection can be inferred. Then, the identified broken road segments connecting the intersection can be extended to the corresponding intersection points to complete the road topology construction and form a complete road network.

[0019] Furthermore, step 3, the intersection turning relationship detection, specifically includes:

[0020] Step 3.1: Construct a Delaunary triangulation network for the identified road network intersections. Then, identify half of the shortest side of the triangulation network as the area of ​​influence of the intersection. Finally, identify the road segments that intersect with the area of ​​influence of the intersection as the road segments associated with the intersection, which serve as the basis for determining the turning relationship.

[0021] Step 3.2: There are two types of situations between the associated road segments: mutual passage or one-way passage. Therefore, the passage trajectory of the associated road segments is first identified. Then, based on the passage trajectory identification, the time difference between the same trajectory points within the influence area of ​​the non-study intersection endpoint of the associated road segments is statistically analyzed to determine the passage status between the associated road segments.

[0022] Step 3.3: Treat the associated road segments as vectors and determine the turning relationship based on the road segment turning rules and turning angle relationships.

[0023] Furthermore, the specific implementation method of step 3.2 is as follows:

[0024] (b1) Identification of travel trajectories of related road segments: The trajectories of the two related road segments within the influence area of ​​the non-study intersection endpoint and the study intersection endpoint are superimposed to obtain the travel trajectories of the two related road segments;

[0025] (b2) Identification of traffic status of associated road segments: After identifying the traffic trajectory of associated road segments, intersect it with the buffer zone of the non-study intersection endpoint of the associated road segment to obtain the traffic trajectory of the non-study intersection endpoint of the associated road segment. Subtract the timestamps of the two traffic trajectories of the non-study intersection endpoints of the associated road segments and determine the traffic status of the associated road segments based on the design rules, which are shown in Table 1. If the absolute value of the time difference satisfies the formula (1), it is considered as a noisy turn and is eliminated.

[0026] |T2 - T1| > (Length(L1) / V1 + Length(L2) / V2), Formula (1), where L1 is the associated road section corresponding to the minuend; L2 is the associated road section corresponding to the subtrahend; T1 is the time stamp corresponding to the passing trajectory point at the non - research intersection end of the L1 road section; T2 is the time stamp corresponding to the passing trajectory point at the non - research intersection end of the L2 road section; length() is the length calculation function; V1 is the average speed of the passing trajectory point at the non - research intersection end of the L1 road section; V2 is the average speed of the passing trajectory point at the non - research intersection end of the L2 road section;

[0027] Table 1 Discrimination Rules for the Passing Status of Associated Road Sections

[0028]

[0029] Where T is a positive value as the result, F is a negative value as the result, and C0, C1, and C2 are thresholds.

[0030] Furthermore, the specific implementation of step 3.3 is as follows;

[0031] Set the traffic direction from L1 to L2. Then L1 is the forward road section, represented as a vector L2 is represented as a vector Vector Relative to vector The magnitude of the deviation angle θ can be obtained according to Formula (2);

[0032]

[0033] If the Z - coordinate of the cross - product of the two vectors > 0, then the included angle α of the associated road section is α = θ; if Z < 0, then α = - θ. According to the range of the α angle, judge the turning relationship of the associated road section. The road section turning rules and the turning angle relationship are as follows:

[0034] (c1) If 45° ≤ a ≤ 135°, the road section turning rule is a left turn;

[0035] (c2) If - 135° ≤ a ≤ - 45°, the road section turning rule is a right turn;

[0036] (c3) If - 45° < a < 45°, the road section turning rule is a straight - through;

[0037] (c4) If 135° < a ≤ 180° or - 180° < a ≤ - 135°, the road section turning rule is a U - turn.

[0038] The present invention also provides an urban road network hierarchical generation system considering the turning relationship at intersections, including the following modules:

[0039] The different level road segment identification module is used to classify the trajectory kernel density based on geometric interval classification to achieve different levels of trajectory point kernel density division. Then, a morphological refinement algorithm is used to refine the different levels of trajectory point kernel density, and then different levels of road segments are generated.

[0040] The module for merging and optimizing road segments of different levels is used to merge the identification results of road segments of different levels to obtain road segment node information. Then, based on this, it performs effective identification of intersections using Delanuary triangulation clustering. Furthermore, it conducts road network topology checks based on the intersections to achieve topological connection of broken road segments and form a complete road network.

[0041] The intersection turning relationship detection module is used to determine the intersection influence area based on the identified intersection, then to detect the intersection-related road segments, then to judge the traffic conditions between the related road segments, and finally to identify the turning relationship of the related road segments using turning rules, thus completing the intersection turning relationship detection.

[0042] Furthermore, the different levels of road segment merging and optimization module specifically includes:

[0043] Step 2.1: Construct a Delaunay triangulation network based on the start and end points of the identified road segment, and use this network for intersection identification; the specific implementation method is as follows:

[0044] (a1) First, based on the node-connected road segments, the edges of the node-connected triangular network are removed, and the two triangular network edges with the largest angle to the road segment are retained.

[0045] (a2) Then, the edges of the remaining Delaunay triangulation are superimposed on the M-meter buffer of the identification road network. If the ratio of the overlap length to its length is less than a certain threshold, it is discarded.

[0046] (a3) Then, traverse each node, cluster the continuous nodes connected by triangles into one class, and remove the points that are not connected by triangles as noise until all nodes have been traversed once, and complete the cluster analysis of the nodes. The average value of the node position of each cluster is identified as the candidate position of the intersection.

[0047] (a4) To optimize the candidate intersection location, the candidate intersection location is adjusted to the point with the highest trajectory kernel density within the maximum range of distances from the candidate intersection to its corresponding cluster node, thus obtaining the final intersection location;

[0048] Step 2.2: Based on the intersection identification results and the clustering node information at the intersection, the information of the broken road segments connecting the intersection can be inferred. Then, the identified broken road segments connecting the intersection can be extended to the corresponding intersection points to complete the road topology construction and form a complete road network.

[0049] Furthermore, the intersection turning relationship detection module specifically includes:

[0050] Step 3.1: Construct a Delaunary triangulation network for the identified road network intersections. Then, identify half of the shortest side of the triangulation network as the area of ​​influence of the intersection. Finally, identify the road segments that intersect with the area of ​​influence of the intersection as the road segments associated with the intersection, which serve as the basis for determining the turning relationship.

[0051] Step 3.2: There are two types of situations between the associated road segments: mutual passage or one-way passage. Therefore, the passage trajectory of the associated road segments is first identified. Then, based on the passage trajectory identification, the time difference between the same trajectory points within the influence area of ​​the non-study intersection endpoint of the associated road segments is statistically analyzed to determine the passage status between the associated road segments.

[0052] Step 3.3: Treat the associated road segments as vectors and determine the turning relationship based on the road segment turning rules and turning angle relationships.

[0053] Furthermore, the specific implementation method of step 3.2 is as follows:

[0054] (b1) Identification of travel trajectories of related road segments: The trajectories of the two related road segments within the influence area of ​​the non-study intersection endpoint and the study intersection endpoint are superimposed to obtain the travel trajectories of the two related road segments;

[0055] (b2) Identification of traffic status of associated road segments: After identifying the traffic trajectory of associated road segments, intersect it with the buffer zone of the non-study intersection endpoint of the associated road segment to obtain the traffic trajectory of the non-study intersection endpoint of the associated road segment. Subtract the timestamps of the two traffic trajectories of the non-study intersection endpoints of the associated road segments and determine the traffic status of the associated road segments based on the design rules, which are shown in Table 1. If the absolute value of the time difference satisfies the formula (1), it is considered as a noisy turn and is eliminated.

[0056] |T2-T1|>(Length(L1) / V1+Length(L2) / V2) Formula (1) Where L1 is the associated road segment corresponding to the minuend; L2 is the associated road segment corresponding to the subtrahend; T1 is the timestamp corresponding to the trajectory point of the non-study intersection endpoint of the L1 road segment; T2 is the timestamp corresponding to the trajectory point of the non-study intersection endpoint of the L2 road segment; length() is the length calculation function; V1 is the average speed of the trajectory point of the non-study intersection endpoint of the L1 road segment; V2 is the average speed of the trajectory point of the non-study intersection endpoint of the L2 road segment;

[0057] Table 1 Rules for Determining Traffic Status of Related Road Sections

[0058]

[0059] Where T is the positive value, F is the negative value, and C0, C1, and C2 are thresholds.

[0060] Compared with existing technologies, the advantages and beneficial effects of this invention are as follows: Firstly, this invention uses a decomposition-combination method for road network identification. While ensuring the quality of identification results for road segments corresponding to high-density trajectory traffic, it identifies road segments of other levels, effectively guaranteeing the accuracy and completeness of the overall results. Secondly, it uses the time difference between trajectory points at non-intersection endpoints of intersection-related road segments to determine the traffic status of associated road segments. This not only effectively avoids the influence of noise-induced turning but also significantly improves computational efficiency. The method proposed in this invention is simple, efficient, and accurate, providing a valuable technical reference for the construction of urban navigation road networks. Attached Figure Description

[0061] Figure 1 This is a flowchart of a method for generating a hierarchical urban road network that takes into account the turning relationships at intersections, according to an embodiment of the present invention.

[0062] Figure 2 This is an example of intersection recognition in this invention.

[0063] Figure 3 This is an example of passage through associated road sections in this invention. Detailed Implementation

[0064] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0065] like Figure 1 As shown, the present invention provides a method for generating a hierarchical urban road network that takes into account the turning relationships at intersections, comprising the following steps:

[0066] Step 1, Identification of different levels of road segments: Based on geometric interval classification, the kernel density of the cleaned trajectory is classified into different levels of trajectory point kernel density. Then, a morphological thinning algorithm is used to refine the kernel density of different levels of trajectory points, thereby generating different levels of road segments.

[0067] Step 2, Merging and Optimizing Road Segments of Different Levels: Merge the identification results of road segments of different levels to obtain road segment node information. Then, based on this, perform effective identification of intersections using Delanuary triangulation clustering. Subsequently, conduct road network topology checks based on the intersections to achieve topological connection of broken road segments and form a complete road network.

[0068] Step 3, Intersection Turning Relationship Detection: Based on the identified intersection, the influence area of ​​the intersection is determined, and then the related road segments of the intersection are detected. Subsequently, the traffic conditions between the related road segments are judged, and finally, the turning relationship of the related road segments is identified by the turning rules, thus completing the intersection turning relationship detection.

[0069] Furthermore, the identification of different road segment levels mentioned in step 1 specifically includes:

[0070] Step 1.1: Perform quality cleaning on the trajectory data, including but not limited to location duplication removal, stop point detection, and Delanuary triangulation denoising. Then, perform kernel density analysis on the cleaned trajectory and classify the trajectory kernel density according to geometric interval classification to achieve different levels of trajectory point kernel density (the default is three levels).

[0071] Step 1.2 involves morphological refinement of the kernel density of different levels of trajectory points, yielding the refinement results for first-level, second-level, and third-level trajectory point kernel densities.

[0072] Step 1.3: The refined kernel density results of the first-level trajectory points are identified as first-level road segments. Then, a suitable buffer zone is established using the first-level road segments to erase the refined kernel density results of the second-level trajectory points, resulting in second-level road segments. Finally, a buffer zone is established using the first and second-level road segments as the core to erase the refined kernel density results of the third-level trajectory points, resulting in third-level road segments. Based on the designed distance between road segments, the buffer zone range is set to 75-200m, and erasure tests are conducted. Among the erasure results, the threshold with the best erasure effect is selected as the buffer zone range.

[0073] Furthermore, the merging and optimization of road segments of different levels mentioned in step 2 specifically includes:

[0074] Step 2.1: The identified road segments are broken road segments that are not connected together. Therefore, the road segment nodes are near the intersection and have a clustering phenomenon, which can effectively reflect the intersection information. Delaunay triangulation is an effective tool for expressing the proximity relationship of entities. Therefore, Delaunay triangulation is constructed based on the first and last points of the identified road segments, and intersection identification is performed accordingly.

[0075] Step 2.2: Based on the intersection identification results and the clustering node information at the intersections, the information on broken road segments connecting the intersections can be inferred. Then, the identified broken road segments connecting the intersections are extended to the corresponding intersection points to complete the road topology construction, forming a complete road network.

[0076] Furthermore, the specific implementation method of step 2.1 is as follows:

[0077] (1) First, based on the node-connected road segments, the edges of the node-connected triangular network are removed, and the two triangular network edges with the largest angle to the road segment are retained.

[0078] (2) Then, the edges of the remaining Delaunay triangulation are superimposed on the 50-meter buffer of the identified road network. If the ratio of the overlap length to its length is less than a certain threshold (default 0.9), it is discarded.

[0079] (3) Then traverse each node, group the continuous nodes connected by triangular lines into one category, and remove the points not connected by triangular lines as noise until all nodes are traversed once to complete the clustering analysis of the nodes. The average value of the node positions in each cluster is identified as the candidate position of the intersection.

[0080] (4) To optimize the candidate position of the intersection, adjust the candidate position of the intersection to the highest point of the trajectory kernel density within the maximum range of the distance from the candidate intersection to the nodes in its corresponding cluster to obtain the final position of the intersection.

[0081] Furthermore, the detection of the turning relationship of the intersection described in step 3 specifically includes:

[0082] Step 3.1, construct a Delaunary triangular mesh for the identified road network intersections, then identify half of the shortest side of the triangular mesh as the range of the intersection influence area, and finally identify the road sections intersecting with the intersection influence area as the intersection-associated road sections as the basis for judging the turning relationship;

[0083] Step 3.2, there are two situations of mutual passage or one-way passage between the associated road sections. Therefore, first identify the passing trajectories of the associated road sections, and then, based on the identification of the passing trajectories, count the time difference between the same trajectory points within the influence area of the non-research intersection endpoints of the associated road sections to determine the passing state between the associated road sections;

[0084] Step 3.3, regard the associated road sections as vectors, and judge the turning relationship based on the road section turning rules and the corner relationship.

[0085] For example, if the passage is from L1 to L2, then L1 is the forward road section, represented as the vector L2 is represented as the vector The vector Relative to the vector The magnitude of the deviation angle θ can be obtained according to formula (2).

[0086]

[0087] If the cross product of the two vectors has a coordinate Z > 0, then the included angle α of the associated road sections is α = θ; if Z < 0, then α = -θ. According to the range of the α angle, the turning relationship of the associated road sections can be judged. The road section turning rules and the corner relationship are as follows:

[0088] (1) If 45° ≤ a ≤ 135°, the road section turning rule is a left turn.

[0089] (2) If -135° ≤ a ≤ -45°, the road section turning rule is a right turn.

[0090] (3) If -45° < a < 45°, the road section turning rule is a straight line.

[0091] (4) If 135° < a ≤ 180° or -180° < a ≤ -135°, the turning rule for the road segment is a U-turn.

[0092] Furthermore, the specific implementation method of step 3.2 is as follows:

[0093] (1) Identification of the passing trajectories of associated road segments. Superimpose the trajectories within the influence areas of the non-study intersection endpoints and the study intersection endpoints of the two associated road segments to obtain the passing trajectories of the two associated road segments.

[0094] Taking Figure 3 as an example, the trajectory passing through the non-intersection endpoint A of road segment L1 can be directly identified by recognizing the trajectory points falling within the influence area of endpoint A, and can be marked as T A . However, these trajectories do not necessarily pass through endpoints B and C. Therefore, to further identify the passing trajectories of the associated road segments, it is possible to further intersect with the trajectory points T B within the influence area of endpoint B and the trajectory points T C within the influence area of endpoint C in sequence, and the passing trajectories of L1 and L2 can be obtained.

[0095] (2) Identification of the passing states of associated road segments. After identifying the passing trajectories of the associated road segments, intersect them with the buffer zones of the non-study intersection endpoints of the associated road segments to obtain the passing trajectories of the non-study intersection endpoints of the associated road segments. Subtract the timestamps of the passing trajectories of the non-study intersection endpoints of the two associated road segments, and based on the design rules (as shown in Table 1), determine the passing states of the associated road segments. It should be noted that if the absolute value of the time difference is relatively large (refer to formula 1), it is considered as noise turning and is excluded.

[0096] |T2 - T1| > (Length(L1) / V1 + Length(L2) / V2) Formula (1) where L1 is the associated road segment corresponding to the minuend; L2 is the associated road segment corresponding to the subtrahend; T1 is the timestamp corresponding to the passing trajectory point of the non-study intersection endpoint of road segment L1; T2 is the timestamp corresponding to the passing trajectory point of the non-study intersection endpoint of road segment L2; length() is the length calculation function; V1 is the average speed of the passing trajectory points of the non-study intersection endpoint of road segment L1; V2 is the average speed of the passing trajectory points of the non-study intersection endpoint of road segment L2.

[0097] Table 1 Discrimination rules for the passing states of associated road segments

[0098]

[0099] Where T is a positive value of the result, F is a negative value of the result, and C0, C1, and C2 are thresholds.

[0100] This invention also provides a hierarchical generation system for urban road networks that takes into account the turning relationships at intersections, characterized by comprising the following modules:

[0101] The different level road segment identification module is used to classify the trajectory kernel density based on geometric interval classification to achieve different levels of trajectory point kernel density division. Then, a morphological refinement algorithm is used to refine the different levels of trajectory point kernel density, and then different levels of road segments are generated.

[0102] The module for merging and optimizing road segments of different levels is used to merge the identification results of road segments of different levels to obtain road segment node information. Then, based on this, it performs effective identification of intersections using Delanuary triangulation clustering. Furthermore, it conducts road network topology checks based on the intersections to achieve topological connection of broken road segments and form a complete road network.

[0103] The intersection turning relationship detection module is used to determine the intersection influence area based on the identified intersection, then to detect the intersection-related road segments, then to judge the traffic conditions between the related road segments, and finally to identify the turning relationship of the related road segments using turning rules, thus completing the intersection turning relationship detection.

[0104] The specific implementation methods of each module are the same as those of each step, and will not be described in this invention.

[0105] It should be understood that any parts not described in detail in this specification belong to the prior art.

[0106] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for hierarchical generation of urban road network considering intersection turning relations, characterized in that, Comprising the following steps: Step 1, different level road section identification: based on geometric interval classification, the trajectory kernel density is classified to realize the division of different level trajectory point kernel density, then a morphological thinning algorithm is used to thin the different level trajectory point kernel density, and then different level road section generation is carried out; Step 2, merging and optimization of different level road sections: the different level road section identification results are merged to obtain road section node information, then based on the Delanuary triangulation network clustering, the intersection effective identification is carried out, and then the road network topology checking is carried out based on the intersection, the topological connection of the broken road section is realized, and the complete road network is formed; Step 3, intersection turning relationship detection: based on the identification of the intersection, the intersection influence area is determined, then the intersection associated road section detection is carried out, then the traffic state between the associated road sections is judged based on this, and finally the turning relationship of the associated road sections is identified by using the turning rules, and the intersection turning relationship detection is completed; The intersection turning relationship detection of step 3 specifically comprises: Step 3.1, constructing a Delaunay triangulation network for the identified road network intersection, then identifying half of the shortest side of the triangulation network as the intersection influence area range, and finally identifying the road sections intersecting with the intersection influence area as the intersection associated road sections as the basis for turning relationship judgment; Step 3.2, there are mutual traffic or one-way traffic between the associated road sections, so first the associated road section traffic trajectory is identified, then based on the traffic trajectory identification, the time difference between the same trajectory points in the non-study intersection endpoint influence area range of the associated road sections is counted to judge the traffic state between the associated road sections; The specific implementation mode of step 3.2 is: (b1) associated road section traffic trajectory identification, superimposing the trajectories in the non-study intersection endpoint and the study intersection endpoint influence area range of the two associated road sections to obtain the traffic trajectories of the two associated road sections; (b2) associated road section traffic state identification, after identifying the associated road section traffic trajectory, intersecting with the non-study intersection endpoint buffer of the associated road section to obtain the non-study intersection endpoint traffic trajectory of the associated road section, subtracting the time stamps of the non-study intersection endpoint traffic trajectories of the two associated road sections, and based on the design rules, the traffic state between the associated road sections is judged, and the design rules are as shown in Table 1; if the absolute value of the time difference satisfies formula (1), it is considered as noise turning and is excluded; |T2-T1|>(Length(L1) / V1+Length(L2) / V2) Formula (1) Wherein L1 is the associated road section corresponding to the minuend; L2 is the associated road section corresponding to the subtrahend; T1 is the time stamp corresponding to the non-study intersection endpoint traffic trajectory point of L1 road section; T2 is the time stamp corresponding to the non-study intersection endpoint traffic trajectory point of L2 road section; length() is a length calculation function; V1 is the average speed of the non-study intersection endpoint traffic trajectory point of L1 road section; V2 is the average speed of the non-study intersection endpoint traffic trajectory point of L2 road section; Table 1 Associated road section traffic state identification rule Wherein T is a positive numerical result, F is a negative numerical result, C0, C1 and C2 are threshold values; Step 3.3, regarding the link as a vector, the turning relationship is determined based on the link turning rules and the corner relationship.

2. The urban road network hierarchical generation method considering intersection turning relationship according to claim 1, wherein: The different level link identification in step 1 specifically includes: Step 1.1, quality cleaning of trajectory data, including position repetition elimination, stay point detection, Delanuary triangulation noise reduction, then kernel density analysis of cleaned trajectory, and classification based on geometric interval to classify trajectory kernel density to achieve different level trajectory point kernel density division, default division into three levels; Step 1.2, morphological thinning of different level trajectory kernel density, which can respectively obtain first level trajectory point kernel density thinning result, second level trajectory point kernel density thinning result, and third level trajectory point kernel density thinning result; Step 1.3, identifying the first level trajectory point kernel density thinning result as a first level link, then establishing a suitable buffer zone based on the first level link to erase the second level trajectory point kernel density thinning result, obtaining the second level link, and finally establishing a buffer zone based on the first and second level links to erase the third level trajectory point kernel density thinning result, obtaining the third level link.

3. The urban road network grading generation method considering intersection turning relationship according to claim 1, wherein: The different level link merging and optimization in step 2 specifically includes: Step 2.1, constructing a Delaunay triangulation based on the identified link head and tail points, and identifying intersections accordingly; the specific implementation is as follows: (a1) first, based on node connection link, remove node connection triangulation edge, and keep the two triangulation edges with the largest angle with the link; (a2) then, superimpose the remaining Delaunay triangulation edge with the identified link M buffer zone, and if the overlap length is less than a certain threshold compared to its length, remove it; (a3) then, iterate through each node, and cluster the continuous nodes connected by triangulation lines into a class, and remove the points without triangulation line connection as noise, until all nodes are iterated, complete the cluster analysis of nodes, and the average position of each class of nodes is identified as the intersection candidate position; (a4) to optimize the intersection candidate position, adjust the intersection candidate position to the highest trajectory kernel density point within the maximum distance range of the candidate intersection to the corresponding cluster node distance, to obtain the final intersection position; Step 2.2, based on the above intersection identification result and the cluster node information at the intersection, the broken link information connecting the intersection can be inferred, and then the identified broken link connecting the intersection is extended to the corresponding intersection to complete the road topology structure construction, forming a complete road network.

4. The method of claim 1, wherein the method further comprises: The specific implementation of step 3.3 is as follows: If L1 is set to pass to L2, L1 is the forward route section, expressed as a vector , L2 is expressed as a vector , the vector is offset from the vector by an angle , the size of which can be obtained according to formula (2); Formula (2) If the coordinate Z>0 of the two vector difference products ( ), the included angle of the associated road segment is = , if Z<0, then = , according to the range of angle, the turning relationship of the associated road segment is judged, and the road segment turning rules and the turning angle relationship are as follows: (c1) If 45°≤ If the angle is ≤135°, the turning rule for this road segment is left turn; (c2) If -135°≤ If the angle is ≤-45°, the turning rule for this road segment is a right turn; (c3) If -45° < If the angle is less than 45°, the turning rule for that road segment is to go straight. (c4) If 135° < ≤180° or -180°≤ If the angle is less than -135°, the turning rule for that road segment is a U-turn.

5. A system for hierarchical generation of urban road network considering intersection turning relations, characterized in that, It includes the following modules: A different level link identification module for classifying trajectory kernel density based on geometric interval to achieve different level trajectory point kernel density division, then using morphological thinning algorithm for different level trajectory point kernel density thinning, and then developing different level link generation; A different level link merging and optimization module for merging different level link identification results to obtain link node information, then based on Delanuary triangulation clustering to effectively identify intersections, and then based on intersections to develop road network topology checking, realize topology connection of broken links, and form a complete road network; The intersection turning relationship detection module is used for intersection influence area determination based on intersection recognition, and then for intersection associated road section detection, and then for associated road section traffic state identification, and finally for turning relationship identification of the associated road section based on turning rules, to complete the intersection turning relationship detection. The intersection turning relationship detection module specifically includes: Step 3.1, constructing a Delaunay triangular network for the identified road network intersection, then identifying half of the shortest side of the triangular network as the intersection influence area range, and finally identifying the intersection associated road section intersecting the intersection influence area as the intersection associated road section as the basis for turning relationship judgment; Step 3.2, there are mutual traffic or one-way traffic between the associated road sections, so first the associated road section traffic trajectory is identified, and then based on the traffic trajectory identification, the time difference between the same trajectory points in the non-research intersection endpoint influence area range of the associated road section is counted to determine the traffic state between the associated road sections; The specific implementation of step 3.2 is: (b1) Associated road section traffic trajectory identification, superimposing the trajectories in the non-research intersection endpoint influence area range of the two associated road sections to obtain the traffic trajectories of the two associated road sections; (b2) Associated road section traffic state identification, after identifying the traffic trajectories of the associated road sections, intersecting the non-research intersection endpoint buffer zone to obtain the non-research intersection endpoint traffic trajectory of the associated road section, subtracting the time stamps of the non-research intersection endpoint traffic trajectories of the two associated road sections, and based on the design rules to determine the traffic state of the associated road section, the design rules are shown in Table 1; if the absolute value of the time difference satisfies formula (1), it is considered as noise turning and is excluded; |T2-T1|>(Length(L1) / V1+Length(L2) / V2) Formula (1) Where L1 is the associated road section corresponding to the minuend; L2 is the associated road section corresponding to the subtrahend; T1 is the time stamp corresponding to the non-research intersection endpoint traffic trajectory point of the L1 road section; T2 is the time stamp corresponding to the non-research intersection endpoint traffic trajectory point of the L2 road section; length() is a length calculation function; V1 is the average speed of the non-research intersection endpoint traffic trajectory point of the L1 road section; V2 is the average speed of the non-research intersection endpoint traffic trajectory point of the L2 road section; Table 1 Associated road section traffic state identification rules Where T is a positive numerical result, F is a negative numerical result, C0, C1 and C2 are threshold values; Step 3.3, regarding the associated road section as a vector, and determining the turning relationship based on the road turning rules and the corner relationship.

6. The urban road network grading generation system considering intersection turning relationship according to claim 5, characterized in that: The different level road section merging and optimization module specifically includes: Step 2.1, constructing a Delaunay triangular network based on the identified road section head and tail points, and identifying intersections based on the triangular network; the specific implementation is: (a1) First, based on the node connection road section, the node connection triangular network edge is removed, and the two triangular network edges with the largest angle with the road section are retained; (a2) Then, the edges of the remaining Delaunay triangular network are superimposed with the identified road network M buffer zone, and if the overlap length is less than a certain threshold value, the edge is removed; (a3) Further, each node is traversed, and the continuous nodes connected by triangular lines are clustered into a class, and the points without triangular line connection are removed as noise, until all nodes are traversed once, the clustering analysis of the nodes is completed, and the average value of the node positions of each cluster is identified as the intersection candidate position; (a4) In order to optimize the intersection candidate position, the intersection candidate position is adjusted to the highest point of the trajectory core density within the maximum range of the distance from the candidate intersection to the corresponding cluster node, and the final intersection position is obtained; Step 2.2, based on the above intersection recognition results and intersection clustering node information, the broken link information connecting the intersections can be inferred, and then the identified broken link connecting the intersections Extension The road topology structure is completed to the corresponding intersection, and a complete road network is formed.

Images